Artificial intelligence (AI), particularly machine learning (ML), is already transforming transport, telecommunications and retail. Globally, the top ten patent applicants and proprietors for AI/ML related-inventions include Microsoft, IBM, Siemens, Google, Zongcheng Li, State Grid Corporation of China, Samsung, Qualcomm, Hitachi and Panasonic.
AI and ML have the promise of revolutionising healthcare. Indeed, AI in healthcare is growing at an accelerating rate, with the market expected to be valued at US$6.6 billion by 2021. Applications of AI in the healthcare sector are far reaching and include drug discovery, diagnosis, treatment, assistance during surgery as well as record management.
With AI and ML set to disrupt the healthcare industry and provide an immense opportunity for innovation, we analyse the intellectual property (IP) landscape in the sector.
Technology giants enter the stage
Notably, AI in healthcare is at the confluence of traditional healthcare organisations, including pharmaceutical and medical device innovators, and the AI/ML leaders, including the technology giants of Facebook, Apple, Microsoft, Google and Amazon. So what are these AI/ML leaders doing in healthcare?
Between 2013 and 2017, Google, Microsoft and Apple filed between them more than 300 patent applications related to healthcare. Ahead was Google with 186 patent applications, Microsoft with 73 and Apple at 54. While Amazon had not published any healthcare related patent applications in this timeframe, their partnership with Berkshire Hathaway and JP Morgan may lead to patent filings in this sector in the future.
The surge in filings of patent applications by companies that have not traditionally been associated with the healthcare sector demonstrates the opportunities for disrupting the healthcare industry with new technologies.
Amongst the five technology giants, Google is firmly positioning itself as the leader in AI/ML. Already ubiquitous online, Google (used herein synonymously with Alphabet) has extended machine learning from shopping suggestions to disrupting healthcare. Google is not only establishing AI/ML research centres globally and developing its own hardware for AI/ML processing but also investing and acquiring emergent AI/ML organisations. DeepMind was acquired by Google and is dedicated to AI research. DeepMind is concerned particularly with data, for example as generated by wearables and from imaging devices such as MRI and CT. Verily Life Sciences partners with existing healthcare institutions to improve healthcare via analytics tools, interventions and research. Projects at Verily include detecting diabetes through wearables, combating presbyopia, identifying indicators of Parkinson’s disease onset and understanding progression of multiple sclerosis. Calico is dedicated to controlling ageing and age – related diseases.
Application of machine learning in healthcare has been enabled particularly through partnerships and collaborations between technology-focused companies and healthcare organisations. For example IBM Watson has collaborated with both Novartis and MIT while Berg Health and Sanofi and Numerate and Takeda have also partnered.
Phillips, for example, has developed clinical decision support systems for both diagnosing and treating cancers as well as neurological disorders, such as Alzheimer’s. Through use of image processing, GE healthcare has advanced also diagnosis in oncology. Siemens is leading decision support systems for distinguishing benign and malignant lesions for breast cancer.
Leaders in AI/ML in healthcare
The following companies are emerging as leaders in AI and ML for healthcare applications:
Drug discovery: IBM Watson, Google Deep Mind, Benevolent AI, Berg Health, Exscientia, Numerate, Atomwise, twoXAR, Pharma AI, Meta, Globavir, CloudPharmaceuticals, Insilico Medicine, GNS Healthcare, Recursion Pharmaceuticals Whole Biome.
Diagnostics: IBM Watson, Google Deep Mind, Proscia, Path AI, Babylon Health, Cenlitic, deep genomics, Entopsis, Pathway Genomics, Freenome.
Medical imaging: Butterfly network, Arterys, Visexcell, Zebra Medical Vision, CureMetrix, Infervision, Nirami, Enlitic, Recursion, Qure.ai, Clarview Diagnostics.
How to obtain patent protection for AI and ML inventions at the European Patent Office
Given the emergence of AI and ML in healthcare applications, how can companies protect their ideas?
Although there are some challenges to protecting AI inventions, it is possible to obtain patent protection in this fast developing area. One of the main features that is assessed during the examination of applications at the European Patent Office (EPO) is the technical character of an invention. Therefore, applications in the AI/ML space need to be drafted carefully to comply with the legal provisions in Europe and elsewhere.
A number of cases have been decided by the Boards of Appeal, the first and final judicial instance in the procedures at the EPO which illustrate the hurdles in protecting innovations relating to AI and ML.
For example, at the EPO, features of an algorithm underlying a computer-implemented method, falling under the exclusions of Article 52(2) and (3) EPC, forbidding patenting of a computer program “as such”, provide a technical contribution only to the extent that they interact with the technical subject matter of a claim for solving a technical problem (T 154/04). A technical interaction may be present if technical considerations motivating the algorithm’s design can be identified that make the algorithm suitable for being performed on a computer and that ‘go beyond merely finding a computer algorithm to carry out some procedure’ (T 1538/09). However, simply improving algorithmic efficiency has been assessed by the Boards to not provide a technical effect in support of inventiveness (T 1784/06, T 42/10, T 1370/11).
In T 1510/10, the Board of Appeal had to consider whether using machine-learning algorithms could contribute to inventive step. This case concerned ranking information, particularly live web applications, based on interest and/or importance. The Board highlighted that the claimed subject matter failed to define any particular method of machine learning – not even one was described in the application. Rather, machine learning was presented in the application as known. Thus, the Board decided that ‘no inventive step can derive just from the use of machine learning.’ The Appeal was dismissed.
In T 1285/10, grant of a patent was allowed following remittal to the department of first instance, but not because of the artificial intelligence routines that the claimed method used. This case related to an artificial intelligence system for genetic analysis, claiming a method for diagnosing and recommending treatment for a physiological condition. Three of the five claimed method steps related to handling hybridization data of an array of oligonucleotides of about 20 mer to about 25 mer or peptide nucleic acid probes. Particularly, step (iv) required using artificial intelligence routines to determine the most likely pathological or physiological conditions suggested by comparative analysis of hybridization profiles. However, this use of AI was not at issue before the Board of Appeal since it was common ground that use of AI generally was already known. Rather, the case turned on the type of data. While the Board was only reviewing the first instance decision on added matter and sufficiency, it did not come to a decision on inventive step, but made an obiter dictum observation that it considered the claims of the requests to be obvious in the light of the prior art. After remittance, the department of first instance decided that the auxiliary request considered in the Appeal was inventive, in view of its use of hybridization information from an array of peptide nucleic acid probes, and a patent granted, published as EP 1222602B1.
So when considering inventions that use artificial intelligence or machine learning, wherein lies the invention? Following T 1510/10 and T 2418/12, the invention should not be in the use of AI or ML per se, as these fail to provide non-excluded technical contributions even if they do give novelty. Rather, the invention should be in the non-excluded technical solution to a technical problem – as apparent in T 1285/10 and exactly as according to the established case law for assessment of patentability of computer-implemented inventions at the EPO.
For applicants, patentees and opponents alike, these decisions help in assessing whether a claimed invention, that uses AI or ML, will succeed or fail. AI or ML may bring novelty, but AI or ML per se does not give the non-excluded technical contribution necessary for an inventive step. At minimum, the claimed invention should provide a technical effect in the real world that is more than simply a way of speeding up arriving at the solution to a problem. Achieving the technical effect should not be reliant solely on the AI or ML, both being arguably well-known and having expected outcomes, at least in a sense of providing improvements. Nevertheless, where AI or ML is used, the application should comprehensively describe the detailed implementation, functionally and structurally – even if only for sufficiency or a view towards the US patent office requirements for patentability.
Furthermore, the recent update to the Guidelines for Examination in the EPO (valid from 01 November 2018) provides examples of inventions that overcome the exclusions and that are patentable at the EPO. Selected examples listed in the Guidelines of Examination of technical purposes served by a mathematical method include providing a genotype estimate based on an analysis of DNA samples, as well as providing a confidence interval for this estimate so as to quantify its reliability. For inventions based on AI and ML, the Guidelines for Examination also include patentable examples of technical applications and technical implementations, including use of a neural network in a heart monitoring apparatus for the purpose of identifying irregular heartbeat.
Recent granted European patents related to AI/ML in healthcare
EP2895048B1 (07 November 2018) from Verily Life Sciences LLC relates to a system for sensing a blink of an eyelid using a contact lens. The system may use classification systems (for example, support vector machines, neural networks, expert systems, Bayesian belief networks, fuzzy logic or data fusion engines) to infer actions. Granted claim 1 recites:
1. A device (100), comprising:
- a contact lens (110) comprising:
- a substrate;
- a plurality of sensors (230) disposed on or within the substrate;
- a control circuit (220) disposed on the substrate and coupled with the sensors, the control circuit comprising:
- a processing component (255) configured to:
- obtain respective state information associated with one or more of the plurality of sensors (230), where in the respective state information indicates whether an associated sensor (230) is covered by an eyelid (150), and wherein respective state information associated with respective sensors (230) is uniquely identifiable by the processing component (255); and
- determine at least one of a blink of the eyelid (150), a position of the eyelid (150), or an orientation of the contact lens (110) based on the state information
- characterized in that the plurality of sensors (230) are arranged at known locations of the contact lens (110) relative to each other and the processing component (255) is further configured to:
- determine an order in which the one or more sensors (230) are covered or uncovered by the eyelid (150); and
- determine the orientation of the contact lens (110) based on the known locations of one or more sensors (230) determined to be covered by the eyelid (150) and the determined order in which the one or more sensors (230) are covered or uncovered by the eyelid (150).
EP3148408B1 (21 March 2018) from Microsoft Technology Licensing LLC relates to adaptive lifestyle metric estimation, using a machine learning approach for accurate ambulatory energy expenditure estimation. Granted claim 1 recites:
1. A wearable electronic device (10) comprising:
- a processor (60); and
- a sensor system providing inputs to the processor (60), the sensor system including a first sensor (62) and a second sensor (64), the first sensor (62) being a heart rate sensor that consumes a greater amount of power when in operation than the second sensor (64), which is an accelerometer;
- wherein the processor (60) is configured to operate in a high power mode in which both sensors (62, 64) are operational and a low power mode in which the second sensor (64) is operational and the first sensor (62) is not operational;
- wherein, in the high power mode, the processor (60) is configured to compute a calorie expenditure about a user for a first time period based on first data from the first sensor (62); and
- wherein, in the low power mode, the processor (60) is configured to compute the calorie expenditure for a second time period based on:
- second data from the second sensor (64); and
- the first data and/or a value derived from the first data (68),
- wherein the value derived from the first data (68) comprises a first value and a second value, the first value being a resting heart rate and the second value being a heart rate recovery rate.
EP2922025B1 (07 November 2018) from Samsung Electronics Co Ltd relates to an apparatus and method for visualising anatomical elements in a medical image. Granted claim 1 recites:
1. An apparatus (30, 300) for visualizing anatomical elements in a medical image (10), the apparatus (30, 300) comprising:
- a display;
- a memory configured to store instructions therein; and
- at least one processor, upon execution of the instructions, configured to:
- receive a medical image (10);
- detect a plurality of anatomical elements from the medical image (10);
- verify a location of each of the plurality of anatomical elements based on anatomical context information (37) comprising location relationships (372) between the plurality of anatomical elements, and adjust the location relationships between the plurality of anatomical elements detected by the at least one processor, based on the anatomical context information (37); and
- combine information related to the verified and adjusted plurality of anatomical elements with the medical image (10); and
- control the display to display the medical image combined with the information on the display;
CHARACTERISED IN THAT
- the at least one processor is further configured to verify whether a region of interest, ROI, detected from the medical image (10) is a lesion based on a lesion detection probability of an anatomical element, of which the location is verified and the location relationship with other detected anatomical elements is adjusted, in which the ROI is located;
- the medical image (10) is one of a plurality of continuous frames; and
- the anatomical context information (37) further comprises adjacent image information comprising location information of anatomical elements detected in an adjacent frame.
EP3004892B1 (10 October 2018) from Nestec SA relates to a method for diagnosing irritable bowel syndrome, using a statistical algorithm such as a random forest, classification and regression tree, boosted tree, neural network, support vector machine, general chi-squared automatic interaction detector model, interactive tree, multiadaptive regression spline and/or machine learning classifier. Granted claim 1 recites:
1. A method for aiding in the diagnosis of irritable bowel syndrome (IBS) and/or a clinical subtype thereof in a subject, said method comprising obtaining at least two of the following (a) through (f) scores:
(a) detecting in a sample obtained from said subject the level of at least one bacterial antigen antibody marker to obtain a microbiome score;
(b) detecting in said sample the level of at least one mast cell marker to obtain a mast cell score;
(c) detecting in said sample the level of at least one inflammatory cell marker to obtain an inflammatory score;
(d) detecting in said sample the level of at least one bile acid malabsorption (BAM) marker to obtain a BAM score;
(e) detecting in said sample the level of at least one kynurenine marker to obtain an oxidative stress score;
(f) detecting in said sample the level of at least one serotonin marker to obtain a serotonin score;
(g) applying a statistical algorithm to the at least two of said microbiome score, said mast cell score, said inflammatory score, said BAM score, said oxidative stress score, and said serotonin score to obtain a disease score; and
(h) determining a diagnosis of IBS in said subject based on a statistical algorithm that generates a probability of having IBS based the disease score and a diagnostic model comprising at least two of a microbiome score, a mast cell score, an inflammatory score, a bile acid malabsorption score, an oxidative stress score, a serotonin score and combinations thereof from a retrospective cohort.
What does the future hold?
With the growing importance of AI and ML in the healthcare sector, we expect to see a further rise in patent applications in this field, in particular with more and more technology companies emerging as stakeholders. AI can be a game changer to how patients are diagnosed and treated. As we are moving into new territory, companies need to protect their intellectual property assets to remain competitive. With challenges in how these types of inventions can be protected, creative and thoughtful strategies are key when it comes to IP protection. AI also provides an opportunity for more collaborative innovation and patent protection strategies.
If your business depends on technology based on AI and ML, we can secure the intellectual property (IP) protection you need, so you can maintain your exclusivity. In addition, we can define a clear path for you through the IP surrounding your technology, so you can continue and grow your business.
Written by Howard Read, Associate at Appleyard Lees. This article was first published on www.appleyardlees.com on 12 November 2018 ahead of BioBeat 2018.
Feature image © k_yu – stock.adobe.com